Smoke movement for sprinklered fires

Lougheed, G.D.; McCartney, C.; Taber, B.C.

Posting courtesy American Society of Heating, Refrigerating and Air-ConditioningEngineers, Inc. www.ashrae.org Originally published in ASHRAE TransactionsDroits d'auteur : affich avec la permission de l'American Society of Heating,

Refrigerating and Air-Conditionning Engineers Inc. www.ashrae.orgPubli l'origine dans les transactions ASHRAE.

www.nrc.ca/irc/ircpubs

NRCC-43138

DA-00-6-1 (RP-976)ABSTRACTThis paper presents the initial results of a project initiated

by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) and the NationalResearch Council of Canada to investigate smoke movementresulting from a sprinklered fire in a communicating space intoan adjacent large open area such as an atrium or retail mall.Recent research on the interaction of sprinkler spray with asmoke layer is also reviewed. In addition, information in theliterature from full-scale fire tests of mercantile and officeoccupancies is discussed.

As part of the joint project, a large-scale test facility wasestablished to investigate smoke flow for sprinklered fires. Thisfacility is described in the paper. The results of steady-state firetests using a propane burner system are discussed. These testsindicate that two smoke flow regimes can occur depending onthe fire size. For fires with low heat release rates, the smoketemperature was uniform over the height of the compartmentopening and was near ambient. Under this condition, thesmoke was nonbuoyant and accumulated near the opening.With higher heat release rates, a two-zone air and smoke flowregime resulted. The smoke exited the compartment in the hotupper layer. The fire test results were used to determine approx-

Gary D. Lougheed, Ph.D. Cam McCMember ASHRAEASHRAE Transactions: Symposia(NFPA 1995; Klote and Milke 1992; Hansel and Morgan1994; BOCA 1996; ICBO 1994). Although there are differ-ences in the equations used for designing an atrium smokemanagement system for a fire on the atrium floor, the generalapproach for atrium smoke management with this fire scenariois similar in each. The atrium smoke management systems relyeither on the use of smoke filling the atrium space or the useof smoke ventilation from the hot upper layer to provide a clearheight in the atrium for a time period sufficient for occupantsto respond and evacuate.

Protecting the occupants of a building from the adverseeffects of smoke in the event of a fire is one of the primaryobjectives of any fire protection system design. Achieving thisobjective becomes more difficult when dealing with very largespaces, such as an atrium or a mall, where a large number ofoccupants may be present and the compartment geometriesmay be complex. Because of these difficulties, model buildingcodes place restrictions on the use of such spaces in buildings.Some of the requirements that are commonly applied in codesinclude:

the installation of automatic sprinklers throughout thebuilding,

artney Bruce C. TaberINTRODUCTIONDesign information concerning smoke management for

atria and malls is provided in codes and engineering guides

The requirements for automatic fire suppression andcontrolling the fuel provide methods to limit the fire size andthus reduce smoke production.

Gary D. Lougheed is a senior research officer and Cam McCartney and Bruce C.Taber are technical officers in the Fire Risk ManagementProgram, Institute for Research in Construction, National Research Council of Canada, Ottawa, Ontario.imate limits for the smoke flow regimes.Preliminary tests were conducted to investigate the effec-

tiveness of opposed airflow systems in limiting smoke flowbetween the test compartment and the adjacent area. Initialresults are also provided for this portion of the investigation.

limitations on the amount of combustible materialseither used in the construction of the building or locatedon the floor of the atrium, and

the provision of smoke management systems to main-tain tenable conditions in egress routes.Smoke Movement for Sprinklered Fires605

Frequently, in buildings with atria or malls, the commu-nicating spaces (shops, walkways, offices, etc.) open onto theinterconnected floor space. There is design guidance providedin engineering guidelines such as NFPA 92B (NFPA 1995) tomanage smoke from a fire that originates in a communicatingspace. One approach is to use the atrium smoke managementsystem to remove any smoke that enters the large, open spaceto limit the depth of smoke accumulation or delay smoke fill-ing. An alternative approach is to prevent smoke originating inthe communicating space from propagating into the largespace by using physical barriers or opposed airflow. The limit-ing average velocity required for an opposed airflow systemcan be calculated using equations developed by Heskestad(1989), which are included in NFPA 92B.

It is assumed in NFPA 92B that it will not be possible tomanage smoke within the communicating space without theuse of physical barriers to limit smoke movement or methodsto limit smoke production, such as controlling the fuel or usingautomatic sprinklers. Although automatic fire suppression isfrequently used to reduce smoke production from fires in acommunicating space, there are limited data available on thepotential size of sprinklered fires and the transport of smokecooled by sprinklers. Data such as that produced by Liu (1977)indicates that the smoke from a sprinklered fire may not bebuoyant. There are also indications that a sprinklered fire canproduce sufficient smoke to result in obscurations that exceednormally accepted tenability limits in the compartment of fireorigin (Lougheed 1997).

In 1997, a joint research project was initiated between theAmerican Society of Heating, Refrigerating and Air-Condi-tioning Engineers (ASHRAE) and the National ResearchCouncil of Canada (NRC) to investigate smoke movementfrom a sprinklered fire in a compartment that is allowed topropagate into a large open area (ASHRAE Research ProjectRP-976). This paper provides initial results from this project,which include a review of previous research on the interactionof sprinkler spray with a smoke layer, heat release rate fromsprinklered fires, and smoke movement for sprinklered fires.The large-scale test facility set up to investigate smoke move-ment from a sprinklered compartment into a large open area isdescribed. Results of tests to determine smoke flow for suchfires with and without opposed airflow are discussed.

REVIEW OF SPRINKLERED FIRESA number of studies have investigated the effects of sprin-

klers on the smoke produced by a fire. This includes studies todetermine the effects of the sprinkler spray on the hot smokelayer, the heat release rate, and other parameters for sprin-klered fires and smoke movement resulting from a sprinkleredfire. In this section, the research in each of these areas issummarized.

Sprinkler Interaction with the Smoke LayerMuch of the research on the effects of sprinklers on the606

hot smoke layer has investigated the effects between sprin-klers and smoke venting in warehouse and industrial applica-tions. McGrattan et al. (1998) and Hinkley et al. (1993a,1993b) summarize research in this area.

Numerical models were developed for the interaction ofsprinkler spray with the hot gases in the smoke layer and in thefire plume (Bullen 1974; Morgan 1979; Alpert 1985; Heske-stad 1991; Cooper 1991; Chow and Fong 1991; Forney andMcGrattan 1995). The models generally take into account theconvective cooling of the hot layer and the entrainment of hotair in the sprinkler spray.

The research indicates that, prior to control of the fire bythe sprinklers, the hot gases in the smoke layer can beentrained by the sprinkler spray, resulting in the smoke layerpenetrating into the cold lower layer. However, if the temper-ature of the induced gas flow is above the temperature of thecold layer, it will experience a buoyant force, resulting in areversal of the downward flow (Heskestad 1991). Thus, if theconvective cooling by the sprinklers is small relative to theheat content in the hot upper layer, the hot layer will remainbuoyant. However, once the sprinklers gain control of the fire,the temperature in the upper layer decreases, leading to a deep-ening of the smoke layer (Morgan 1979). With furtherdecreases in temperature, the buoyancy of the upper layer willdecay and the smoke will be transported to the floor and even-tually dispersed throughout the test volume (Heskestad 1991).

A limited number of experimental studies wereconducted to investigate the cooling of smoke by sprinklerwater spray. Liu (1977) investigated the cooling produced bya corridor sprinkler system. In these tests, no sprinklers wereincluded in the burn room. The interaction of the water sprayfrom the corridor sprinkler system on the hot gas flow in thecorridor was determined. These tests indicated that the corri-dor sprinkler system was effective in cooling the hot gas flow.In some cases, the smoke temperature exiting the corridor wasless than ambient, resulting in nonbuoyant flow.

Morgan and Baines (1979) conducted tests using an exist-ing large-scale test facility, which represented a part of acovered mall. The facility included a store (fire compartment)and a section of an adjacent mall with the sprinkler installed inthe latter area. As with the investigations by Liu (1977), thesetests determined the heat transfer between the sprinkler sprayand the hot gas flow produced by the fire. The results indicatedthat sprinklers could remove a significant amount of heat andbuoyancy from a hot gas layer.

A series of tests were conducted to investigate the cool-ing of a smoke layer by sprinkler spray for a fire in a compart-ment (You et al. 1986, 1989). These tests were conducted in a3.66 m by 7.32 m by 2.44 m high compartment with a 1.22 mby 2.44 m high opening. A single sprinkler was installed in thetest compartment. The fire source was a heptane spray firewith total heat release rates of up to approximately 500 kW.These tests were used to determine empirical correlations forthe heat absorption by the sprinkler spray and the rate ofASHRAE Transactions: Symposia

convective heat loss through the room opening.

Mawhinney and Tamura (1994) conducted two series oftests using shielded wood crib fires to investigate the interac-tion of sprinklers with the smoke layer in a compartment. Thefirst test series was conducted using a one-story test room andthe second was conducted on the seventh floor of a ten-storytest facility. In the tests in the one-story room, the center of thewood cribs was shielded from direct impingement by waterspray and burning continued in spite of the sprinklers.However, the heat release rate and the temperatures in the testcompartment were reduced, with the amount of reductiondependent on the sprinkler application density. At a density of4.1 L/min per m2, the heat release rate was approximately850-1200 kW and the temperature near the ceiling wasapproximately 150C.

Size of Sprinklered FiresThe performance objective of automatic sprinklers

installed in accordance with NFPA 13 (NFPA 1996) is toprovide fire control that is defined as follows: limiting the sizeof a fire by distribution of water so as to decrease the heatrelease rate and pre-wet adjacent combustibles, while control-ling ceiling gas temperatures to avoid structural damage. Alimited number of investigations have been undertaken inwhich full-scale fire tests were conducted in which the sprin-kler system was challenged but provided this level of perfor-mance. The results of these studies are summarized in thissection.

There are limited full-scale test data available for use indetermining design fire size for sprinklered occupancies.Hansell and Morgan (1994) provide conservative estimatesfor the convective heat release rate based on U.K. fire statis-tics: 1 MW for a sprinklered office, 0.5-1.0 MW for a sprin-klered hotel bedroom, and 5 MW for a sprinklered retailoccupancy. For the latter case, it was assumed that the fire sizeis 3 m by 3 m. These steady-state design fires assume the areais fitted with standard response sprinklers.

Full-scale sprinklered fire tests were conducted for open-plan office scenarios (Madrzykowski and Vettori 1992;Lougheed 1997). These tests indicate that there is an exponen-tial decay in the heat release rate for the sprinklered fires afterthe sprinklers are activated and achieve control. Although apeak heat release rate of approximately 1 MW was measured,the results of these tests indicate that a design fire with asteady-state heat release rate of 500 kW would provide aconservative estimate for a sprinklered open-plan office.

For the open-plan office test (Lougheed 1997), thetemperature in the smoke layer in the compartment wasquickly reduced to less than 100C with the operation of thesprinklers. With a shielded fire scenario used in the testarrangement, the smoke initially remained buoyant aftersprinkler operation. However, as the fire decayed, the smokelost its buoyancy and was mixed throughout the fire compart-ment.

Full-scale fire tests for retail occupancies were conductedASHRAE Transactions: Symposia

in Australia (Bennetts et al. 1997). These tests indicated thatfor many common retail areas (clothing and bookstores), thefire was controlled and eventually extinguished with a singlesprinkler. However, tests simulating areas with shielded fireloads did result in large fires. This includes tests simulating afire in a toy store and in a storage area for a shoe store. Onlyfour active normal response sprinklers were installed for thesetests with a water flow rate of approximately 91 L/min persprinkler with four sprinklers operating.

The toy store had predominantly plastic combustiblesarranged in high shelving. The ignition source was locatedaway from the sprinklers to minimize direct impingement onthe fire during the initial fire growth. With this scenario, thefire continued to spread along the shelving but did not spreadto other racks. The sprinklers limited the temperatures at theceiling and at the edge of the test structure. The temperature ofthe smoke leaving the test area was generally less than 100Ccompared to more than 1000C for the nonsprinklered test.The smoke remained buoyant. The temperature at the ceilingof the test facility was less than 40C compared to 320C forthe nonsprinklered test.

For the tests simulating the shoe store storage area, theshelves were spaced much closer together than for the toystore. The spacing and height of the shelving met Australianstandards. However, based on water spray tests, minimalwater would reach the base of the test arrangement.

In both the sprinklered and nonsprinklered tests, the firetook a considerable time (10 to 30 minutes) to develop. Oncefully developed, the four activated sprinklers were not able tofully control the fire. However, they did substantially reducethe temperatures in the test structure and at the ceiling of themain test facility (145C versus 360C for the nonsprinkleredtest). Also, the temperature in the vicinity of the sprinklers was200C-250C, suggesting that more sprinklers would haveactivated in a real fire situation.

Full-scale fire tests were conducted for a variety of occu-pancies (retail, cellular offices, and libraries) in the U.K.(Ghosh 1997). The retail fire simulations included supermar-ket displays, carpet displays, video stores, and liquor stores.For these tests, the sprinklers controlled the fire. The shelvingsheltered the fuel and the fire could continue to spread alongthe shelves. However, the sprinklers did decrease the temper-ature of the gases, reducing the convective portion of the heatrelease rate from the test area to 30%-50% of the total heatrelease rate compared to 65%-75% for nonsprinklered tests.

The convective heat release rate per unit area for sprin-klered fires varied from 150 kW/m2 to 650 kW/m2. This indi-cates that the use of a constant 500 kW/m2 convective heatrelease rate for sprinklered fires as suggested by Hansell andMorgan (1994) may not always be justified (Ghosh 1997).

Full-scale fire tests were conducted for compact mobilestorage systems used for document storage (Lougheed et al.1994). Such systems are frequently used in a variety of appli-cations including libraries, offices, and hospitals. Informationon tests conducted in 1979 on behalf of the Library of607

Congress is provided in Appendix G of NFPA 910 (NFPA

1991). Subsequent full-scale fire tests were conducted for theLibrary of Congress Archives II and the National Library ofCanada and showed that fires in compact mobile systems aredifficult to extinguish and large quantities of smoke could beproduced even after sprinkler activation (Lougheed et al.1994).

The results of the research summarized in this sectionindicate that, for a fire control situation, the heat release rateis reduced but smoke can continue to be produced. Thetemperature of the smoke is also reduced. The smoke wasbuoyant, at least in the initial stages of tests with larger firesizes. However, once the sprinklers begin to suppress the fire,the smoke can become nonbuoyant and remain at floor level.

Smoke Movement for Sprinklered FiresThere are very limited test data providing information on

smoke movement to areas outside the fire zone for sprinkleredfires. However, some studies do provide useful information.These studies are summarized in this section.

As part of an extended study on fires in patient rooms ofhealth care facilities, 21 full-scale fire tests were conducted atthe U.S. National Institute of Standards and Technology.These tests involved either a mattress with bedding or aclothing wardrobe. Under nonsprinklered conditions, thecombustible clothing wardrobe fire resulted in room flash-over in 120 seconds and rapid development of a smoke layerin the room and adjoining corridor with smoke obscurationand CO concentrations exceeding tenability levels through-out the test area (ONeill et al. 1980).

In subsequent tests with standard 71C pendent sprinklersarranged to provide a 6.9 (L/min)/m2 application density, theceiling gas temperature was lowered. However, the fire couldstill be seen burning inside the wardrobe until smoke obscuredvisibility 60 seconds after the activation of the sprinkler. Also,very high concentrations of CO were measured at the 1.5 mheight throughout the test area with the instantaneous hazard-ous threshold of 1% exceeded in the patient room, the corridor,and the remote lobby area.

In tests with the same wardrobe fire scenario using a stan-dard 71C horizontal sidewall sprinkler arranged to providethe same application density as the pendent sprinkler, betterfire control was obtained. CO concentrations were signifi-cantly lower.

As part of the research on shielded wood crib fires,Mawhinney and Tamura (1994) investigated smoke move-ment in a ten-story test facility set up to simulate a buildingwith zoned smoke control. The smoke control systemprevented smoke spread to adjacent pressurized floors.However, the smoke could spread into the stairshaft if the doorto the fire floor was open. By opening the doors to the stair-shaft on other floors or to the exterior, the stairshaft could bepressurized, reducing the smoke flow into the stairshaft. Also,with the zoned smoke control system shut off, the buoyancypressures were sufficient to cause smoke spread into adjacent608

areas. Lougheed and Carpenter (1997) investigated smokemovement into a corridor and subsequently into a stair shaftusing a ten-story test facility. The fire scenario was similar tothose used in a series of sprinklered open-plan office tests(Lougheed 1997). Differences in sprinkler operation duringthe initial stage of the tests had an impact on the smoke temper-ature in the corridor and thus on the extent and rate at whichsmoke spread in the test facility. Shortly after sprinkler acti-vation, the temperature and CO concentrations measured atvarious locations in the corridor did not vary substantially withheight, indicating minimal smoke stratification once the sprin-klers began to suppress the fire.

Review Summary The available information on the interaction of sprinkler

spray on a smoke layer, the size of sprinklered fires, and smokemovement into adjacent areas for a sprinklered fire werereviewed in this section. Although the research in this area islimited, some general trends can be determined.

1. The size of a fire will be limited by the sprinklers, assumingthe sprinkler system meets its design objective. However,during the fire control phase, a fire can continue to burn,producing smoke. The size of the sprinklered fire dependson the occupancy (the typical fuel load) and the ability ofthe sprinkler spray to reach the seat of the fire (shieldingeffects of furnishing, shelving, etc.). An analysis of U.K.fire statistics indicates that the major effect of sprinklers isto reduce the probability of a sprinklered fire reaching agiven size compared to the equivalent nonsprinklered case(Morgan 1998). For example, approximately 7% of fires insprinklered public-access areas and 13% of fires in sprin-klered non-public areas of retail occupancies will exceed afire damage area of 10 m2.

2. For office occupancies, the fire size in office areas is limitedto 1 MW or less. A steady 500 kW fire would provide aconservative design fire. However, assuming an exponen-tial decay in fire size provides a more accurate representa-tion of the suppression phase. The potential fire size indocument and other storage areas has not been determined.However, there are indications that a substantial sprinkleredfire could develop in such areas (Lougheed 1997).

3. For retail occupancies, the fire size will depend on the useof the space. Based on a review of U.K. fire statistics, aconservative design fire of 5 MW has been used extensivelyfor design purposes. There are, however, indications thatthe 5 MW design fire size is conservative for many appli-cations (Ghosh 1997). Tests simulating fires in clothingstores, which generally constitute a high proportion of thefloor area in modern shopping centers, were suppressed bythe sprinklers (Bennetts et al. 1997). However, there areother areas in retail facilities in which there can be substan-tial fire loads shielded from direct water spray and in whichlarger fires may occur (Ghosh 1997; Bennetts et al. 1997).ASHRAE Transactions: Symposia

Considering the broad range of use of non-public retail

areas, it is difficult to develop a single design fire, andfurther work is required to develop a range of representativedesign fires (Ghosh 1997).

4. During the early stages of a sprinklered fire, the water spraywill cool the hot layer. However, the smoke will generallyremain buoyant until the sprinklers begin to suppress thefire (Heskestad 1991). This smoke will rise above floorlevel once it exits the sprinklered area (Bennetts et al. 1997).

5. With fire suppression, the smoke layer is rapidly cooled andit will lose its buoyancy. During this stage, the smoke willbe distributed over the entire height of the compartment(Heskestad 1991; Lougheed 1997). A similar result can alsooccur if there is cooling by water spray from sprinklers inadjacent areas (Liu 1977). The conditions under which thetransition between buoyant and nonbuoyant smoke flowoccurs was investigated in the full-scale tests discussed inthis paper.

6. The temperature of the smoke outside the fire zone is

tions, the CO concentrations and smoke obscuration inadjacent areas exceeds generally accepted tenability limits(Mawhinney and Tamura 1994; Lougheed and Carpenter1997). There are also indications that the extent and rate ofsmoke spread for sprinklered fires is dependent on thetemperature of the smoke leaving the fire area (Lougheedand Carpenter 1997).

DESCRIPTION OF EXPERIMENT A large-scale test facility was set up to investigate the

impact of sprinklers on smoke movement from a compartmentinto a large adjacent area. This facility was used to conduct aseries of tests with propane-fueled fires to determine theparameters that affect the smoke temperature.The parametersthat determine whether the smoke will be buoyant or nonbuoy-ant were also investigated. Tests were also conducted to deter-mine the ability of opposed airflow systems to limit smokeflow through the opening in the test compartment. Tests simu-lating retail fire situations are planned for the second phase ofdependent on a combination of parameters, such as the heat

release rate of the fire and the sprinkler spray density(Mawhinney and Tamura 1994). However, there are verylimited data available upon which to determine the convec-tive heat content and, thus, the temperature of the smokeleaving the sprinklered area. For design purposes, it can beassumed that the maximum temperature of the smoke exit-ing a sprinklered fire area is approximately the activationtemperature of the sprinklers (Morgan and Gardiner 1990).If the smoke temperature substantially exceeds this temper-ature, further sprinklers will be activated, resulting in acooling of the smoke layer. The various parameters thataffect the smoke temperature exiting the fire zone arediscussed in this paper.

7. There is very limited information available on smokemovement external to the fire zone. Under some test condi-ASHRAE Transactions: Symposia

Figure 1 Large-scale atrium facility and instrumentation (plthe project. The results of these tests will be discussed in asubsequent paper. A description of the large-scale test facilityis provided in this section.

Test FacilityThe test facility setup for this project made use of a large-

scale atrium physical model used for a previous joint projectbetween ASHRAE and NRC (ASHRAE Research ProjectRP-899). Detailed information on the atrium physical modelis provided in the paper by Lougheed et al. (1999). The atriumfacility had a floor area of 13.1 m by 17.2 m and a height of12.2 m.

For the present test series, a test compartment wasconstructed inside the atrium facility. A plan and elevationview of the atrium test facility are shown in Figures 1 and 2. 609

an).

610

The test compartment was a two-story steel structure witha floor area of 5.2 m by 9.2 m and an overall height of 6.4 m.A secondary floor was installed at the 3.39 m height to providea raised test compartment simulating a retail space on an upperlevel of a retail mall. This compartment was used for all thetests discussed in this paper and had a floor to ceiling heightof 3.01 m. Plan and elevation views of the test compartmentare shown in Figures 3 and 4, respectively.

Three walls of the test compartment were lined with 18gauge galvanized steel sheets. The fourth side, located at thecenter of the atrium facility, was open to provide ventilation tothe fire and to allow the smoke produced by the fire to enter thelarge secondary test space. The test arrangement was such that

Figure 2 Large scale atrium facility and instrumentation (ele

Figure 3 Test room and instrumentation (plan).vation).ASHRAE Transactions: Symposia

the centerline of the open wall in the test compartment was atthe center of the atrium facility (Figures 1 and 2).

For most of the tests, the opening was the full height of thecompartment, allowing the smoke produced by the fire to flowalong the ceiling of the compartment and into the adjacent testarea. The opening was 5.0 m wide by 3.01 m high.

Selected tests were conducted with a 0.6 m deep soffitat the top of the opening simulating the case in which thedoor opening between the communicating space and themall does not extend to the ceiling. For these tests, the open-ing was 5.0 m wide by 2.4 m high. The tests with this open-ing arrangement also investigated the effects of thedevelopment of a contained hot layer in the test compart-ment on the smoke exiting the test compartment.

Instrumentation

Thermocouple trees were located in the test compartmentand the atrium facility. The locations of the thermocoupletrees in the atrium facility are shown in Figures 1 and 2. Theplume thermocouple tree (Location 1) and the quarter pointthermocouple tree (Location 2) were used for the testsdiscussed in this paper. The plume thermocouple tree waslocated 0.5 m from the test compartment (Figure 1) and on thecenterline of the opening in the compartment. The quarterpoint thermocouple tree was located at the northeast quarterpoint of the atrium facility. For both thermocouple trees, thespacing between thermocouples was 0.5 m starting at a heightof 0.69 m and 3.2 m above the floor of the main test facility for

Figure 4 Test room and instrumentation (elevation).ASHRAE Transactions: Symposia

were located at the width, the center, and the width of theopening. Six thermocouples spaced at 450 mm were installedon each tree starting at a height of 730 mm above the floor ofthe compartment. The top thermocouples were 2.98 m abovethe floor and approximately 30 mm below the ceiling of thecompartment.

Two thermocouple trees, with six thermocouples on eachtree, were located inside the test compartment (Figures 3 and4). These thermocouple trees were located on the centerline ofthe compartment 3.04 and 6.24 m from the compartmentopening. The thermocouples were spaced at 500 mm starting400 mm above the compartment floor.

Thin metal shields were installed above the thermocou-ples on the trees inside the test compartment and in the open-ing. These shields were 230 mm in diameter and minimizeddirect water spray from the sprinklers reaching the thermo-couples. It was shown in a previous test program (Lougheed1997) that such shields are effective in minimizing the coolingof the thermocouples by the water spray, allowing the temper-ature of the air to be measured.

In addition to the room and opening thermocouple trees,thermocouples were also installed adjacent to each of the foursprinklers in the test compartment (Figures 3 and 4).

Two smoke meters were located in the atrium facility nearthe compartment opening (Figures 1 and 2). These smokemeters were located 3.5 m from the test compartment at thecenterline of the opening and were 1.7 m and 3.7 m above thefloor of the main test facility. They were used to determine ifthe smoke exiting the test compartment was nonbuoyant,resulting in downward smoke flow.

Two video cameras were located in the atrium facility.

inside theted on thempartmentthe plume and quarter point thermocouple trees, respectively.Three thermocouple trees were installed in the opening in

the compartment (Figures 3 and 4). These thermocouple trees

One camera was on the south wall viewing the firetest compartment. The second camera was mouneast wall at the same height as the top of the co611

opening and was used to obtain video records of the smokeplume as it exited the test compartment. Video records werealso taken of smoke conditions at floor level.

Propane Burner Fire SourceA propane burner was used as the fire source for the

majority of the tests discussed in this paper. Propane wasprovided through a T-connection located at the center of a2.7 m long, 50 mm I.D. steel pipe. The pipe was capped atboth ends. Two rows of 70 equispaced 4.7 mm diameterholes were located in the bottom section of the pipe. Thepipe was mounted 410 mm above the floor and was shieldedby a 1.23 m wide by 2.45 m long metal table. The top of thetable was 640 mm above the pipe.

Rotameters were mounted in the propane line, which fedthe propane from the main propane tanks to the burner.Three rotameter sets with ranges of 62-624, 131.1-1311, and245.5-2455 standard L/min were used in series and provided

measurements over the complete range of flow rates used inthe test series. The propane flow rate was used to estimatethe heat release rate using the measured heat of combustionfor the propane.

The propane burner was used to conduct a series ofsteady-state fire tests. For each test, the propane flow rate wasset at an initial level. Sufficient time was allowed for condi-tions in the compartment to come to steady state. The propaneburner was subsequently adjusted to higher flow rates withsufficient time allowed at each step to obtain steady condi-tions. In this way, conditions in the test compartment could bedetermined for several heat releases under the same test condi-tions.

Sprinkler SystemA sprinkler system with four pendent sprinklers was

installed in the test compartment. The location of the sprin-klers is shown in Figure 3. The protection area for each sprin-kler was 12 m2 to simulate a sprinkler system in an ordinaryhazard occupancy. The distance between the two branch lineswas 4.6 m, which is the maximum spacing allowed for an ordi-nary hazard occupancy. The sprinkler spacing on each branchline was 2.6 m. The sprinklers were 2.3 m from the back wallof the test compartment and 1.3 m from the sidewalls.

Tests were conducted with the propane burner at two loca-tions in the test compartment. One series of tests wasconducted with the burner 1.6 m from the back (north) wall ofthe test compartment (Location A in Figure 3). This simulateda fire near a wall with the hot gases exiting through the sprin-kler spray. Tests were conducted with two and four operatingsprinklers.

The second series of tests was conducted with the propaneburner located midway between the two sprinkler branches(Location B in Figure 3). This simulated a fire in the centralarea of a room with the nearest four sprinklers activated.

Tests were conducted with three water flow ratesselected to provide an application density of 4.1, 6.1, and8.1 (L/min)/m2, which covers the design range in NFPA 13(NFPA 1996) for ordinary hazard occupancies. A pressuregauge was installed in the main sprinkler pipe and was usedto monitor the water flow rate during a test.

A thermocouple was installed in the sprinkler pipe and thetemperature of the sprinkler water was recorded throughout atest. The water for the sprinkler system was obtained from acistern located below the test facility. The water temperaturewas constant throughout a test. The water temperature was,however, dependent on the time of year in which the test wasconducted. It varied between 12C and 16C.

Fan System for Opposed Flow TestsTo simulate opposed airflow systems, mechanical

exhaust systems were connected to the back wall of the testcompartment. One portion of the system used two nominal2.8 m3/s fans connected through 460 mm diameter ducts to six612

exhaust inlets in the test compartment. The exhaust inletswere 0.85 m from the sidewalls and were centered 0.29 m,1.34 m, and 2.77 m above the compartment floor. The secondportion of the system used a nominal 7.1 m3/s fan connectedto two 610 mm diameter exhaust inlets located at the center ofthe wall of the compartment. The exhaust inlets were centered1.34 m and 2.67 m above the compartment floor.

Tests were conducted with three fan/inlet arrangements:1. The two small fans operating using the two mid-height

inlets with a total fan capacity of 5.6 m3/s.2. Three fans operating using the highest inlets with a total fan

capacity of 12.7 m3/s.3. Three fans operating using the three mid-height inlets with

a total fan capacity of 12.7 m3/s.

Test ParametersApproximately 50 tests covering 125 test conditions were

conducted with the propane burner fire source to investigatethe impact of various parameters. The parameters included 1. location of the fire (Location A and B in Figure 3),2. heat release rates (150, 200, 250, 300, 400, 500, 600, 750,

1000, 2000, 3000 kW),3. number of sprinklers operating (two and four),4. sprinkler application density (4.1, 6.1, and 8.1 (L/min)/m2),5. with opposed airflow at different flow rates,6. with and without soffit at the top of the compartment open-

ing.

PROPANE BURNER TEST RESULTS

Smoke Flow TestsEach fire test consisted of a series of steady-state fires.

The other parameters, including the sprinkler flow rate,remained constant.

During each stage of the test, the propane flow rate andthus the total heat release rate were maintained steady forsufficient time for the temperature conditions in the testcompartment to stabilize. Figure 5 shows the temperaturemeasured at a thermocouple tree in the compartment open-ing for a typical test. The heat release rates were 1000,2000, and 3000 kW and the sprinkler application densitywas 6.1 (L/min)/m2. With each change in heat release rate,the temperatures measured in the compartment openingquickly adjusted to relatively constant levels. A similarresult was also noted for the temperatures measured insidethe compartment.

The temperature data were used to determine an averagetemperature at each thermocouple location to develop temper-ature versus height profiles for the steady-state conditions.Since there is no provision for conditioning the environmentin the test facility, the ambient temperature varied between10C and 30C depending on the time of year. To simplify thecomparison of temperature profiles between tests, the ambientASHRAE Transactions: Symposia

temperature was subtracted from the measured temperature.

The results were plotted as temperature rise versus heightprofiles.

The temperature profiles obtained at the three-thermo-couple locations in the compartment opening for a series of

Figure 5 Temperature profiles for thermocouples in the compand 3000 kW and a 6.1 (L/min)/m2 sprinkler applicASHRAE Transactions: Symposia

Since the temperature variation across the compart-ment opening was minimal, the profiles at the threecompartment opening locations were averaged to produce aseries of composite temperature rise versus height profiles.Temperature rise profiles are shown in Figures 7-9 for testsconducted with the two locations for the propane burnerand three sprinkler application densities with four operat-ing sprinklers. The heat release rates were 250, 1000, and2000 kW. The temperature profile for tests without sprin-klers is also shown in Figure 7. In addition, the results fortests conducted at 1000 kW with two sprinklers are shownin Figure 8.

As shown in Figures 7-9, the burner location, the numberof operating sprinklers, and the sprinkler application densityimpacted on the temperature rise in the smoke exiting the testcompartment. The variations in the profiles were consistentwith those expected with convective cooling of the smoke bythe sprinkler spray. These include

1. a reduction in temperature for tests with four operatingsprinklers compared to two sprinklers,2. a reduction in temperature for tests with the burner locatedat the back of the compartment compared with the middleof the compartment, and

3. a systematic decrease in the temperature with sprinkler

artment opening for test with heat release rates of 1000, 2000,ation density.tests at 500 kW are shown in Figure 6. The temperatureprofiles are similar at each location except for small varia-tions. These differences are most likely due to the dynamicsproduced in the compartment by the fire and sprinkler spray.

application density.

Although there may be some reduction in the amount ofpropane fuel burned, the test results indicate that suppression613

of the fire by the sprinklers was minimized. By using ashielded propane burner, the smoke flow tests were focused onthe impact of the water spray from the sprinklers on the smokeproduced by the fire. The differences in the temperature riseprofiles are indicative of the changes in convective heat in thesmoke layer exiting the compartment with changes in testparameters.

Temperature rise versus heat release rate at the six eleva-tions in the compartment opening are shown in Figures 10 and11 for sprinkler flow rates of 4.1 and 8.1 (L/min)/m2, respec-tively. In all cases, the burner was located at the back of thecompartment. The test results indicate that there were threesmoke flow regimes with the sprinklered fires:

1. If the convective cooling of the smoke by the sprinklers wassmall compared to the convective heat produced by the fire,the predominant smoke flow is in a hot upper layer. Thedepth of this layer remained relatively constant at approxi-mately 1 m. The temperature in the lower portion of theopening was near ambient. The airflow in this region waspredominantly into the compartment to provide combus-tion air.

614

Figure 6 Temperature profiles at three locations in the comp

Figure 7 Average temperature profiles in opening for tests atASHRAE Transactions: Symposia

artment opening for tests at 500 kW.

250 kW.

ASHRAE Transactions: Symposia

Figure 8 Average temperature profiles in opening for tests at

Figure 9 Average temperature profiles in opening for tests at

Figure 10 Temperature rise vs. heat release rate with 4.1 (L/m 1000 kW.

2000 kW. 615

in)/m2 sprinkler.

2. If the convective cooling of the smoke by the sprinklers washigh compared to the convective heat produced by the fire,a hot upper layer was not formed. The temperature wasuniform over the height of the compartment opening. Thetemperature was slightly above ambient.

3. For intermediate heat release rates, there was a transitionzone in which a hot layer formed at the ceiling and there wasalso a mixture of cold smoke in the lower layer. Withincreasing heat release rate, there was a shift to the two-zone flow regime as the temperatures increased in the upperlayer and decreased to near ambient in the lower portion ofthe compartment.

Using the temperature rise versus heat release rate data,approximate upper and lower limits can be estimated for thethree smoke flow regimes. The results are shown in Table 1.

Figure 11 Temperature rise vs. heat release rate with 8.1 (L/m616

release rate above which the temperature at ceiling levelwas higher than at the lower levels.

TABLE 1 Heat Release Rate Limits Smoke Flow Regimes

Sprinkler Application

Density ((L/min)/m2)

Burner Location

Upper Limit Cold Smoke

Regime (kW)

Lower Limit Two-Zone

Regime(kW)

4.1 Middle

H = height of the opening (m),Tf = temperature of heat smoke (K),To = temperature of ambient air (K).

Preliminary tests were conducted with the fan systemsdescribed in this paper. All the tests were conducted with thefull compartment opening. The estimated average velocitythrough the opening using the two small fans with a totalcapacity of 5.6 m3/s was 0.37 m/s. With the three fans operat-ing with a total capacity of 12.7 m3/s, the estimated velocitywas 0.85 m/s.The flow rates produced using the three-fansystem should be effective for low heat release fires in whichthe temperature rise is less than 20C.

Figure 12 shows the average temperature measured in thecompartment opening for tests conducted at 500 kW. For thesetests, the burner was located at the middle of the compartmentand four sprinklers were used. The results indicate that thesmoke had to be relatively cool for the opposing airflow tocompletely reduce smoke flow through the opening. With

volume space as a result of a sprinklered fire in an adjacentcompartment.

The recent literature on the effects of sprinkler spray onsmoke flow is reviewed. The research in this area is limited,with much of the work related to the interaction between sprin-klers and building vents. However, there were indications thata hot smoke layer will be maintained until the sprinklers beginto suppress the fire. This observation is consistent with theresults obtained in the propane burner fire tests conducted inthe full-scale test facility, which are described in this paper.Specifically, if the heat release rate exceeded a lower limit, thesmoke flow through the compartment opening was predomi-nantly in a hot layer in the upper portion of the opening.

Based on the test results, approximate heat release ratelimits were determined above which hot smoke flow waspredominant for the test conditions. The limits were depen-dent on fire location and sprinkler application density ranging

2

lower sprinkler application densities, there was sufficienttemperature in the hot layer to result in smoke flow through theopening.

There are indications that improved performance wasobtained in tests with the exhaust inlets located in the upperportion of the compartment, resulting in direct exhaust fromthe smoke layer. However, the tests conducted thus far are atthe lower operating limits of such a system. Tests are plannedwith the compartment opening partially blocked to provideincreased airflow through the opening. These tests willprovide a better indication of the range of performance for theopposed air flow system.

CONCLUSIONSThis paper provides initial results of a joint ASHRAE and

NRC research project to investigate smoke flow into a large-ASHRAE Transactions: Symposia

Figure 12 Temperature profiles in opening for tests at 500 kWbetween approximately 150 kW to 750 kW for 4.1 (L/min)/mand 8.1 (L/min)/m2, respectively.

With increasing heat release rates, there was a steadyincrease in the temperatures measured in the upper layer.However, for high heat release rates, the temperature near theceiling exceeded 100C, and further sprinklers would beexpected to operate in a real fire, cooling the upper layer(Morgan and Gardiner 1990). The limits within which furthersprinklers might operate in the test configuration ranged fromapproximately 1500 kW for the low sprinkler applicationdensity to approximately 2500 kW for the high sprinkler appli-cation density. This is within the range of fire sizes that couldbe expected in retail occupancies with high fire loads (Ghosh1997; Bennetts et al. 1997). Further fire tests are planned toassess the smoke flow with realistic fire scenarios and thepotential smoke filling in the adjacent large space.617

with opposed air flow.

For low heat release rate fires (

tain fire test projectLarge scale experiments andmodel development. Quincy, Mass.: National Fire Pro-tection Research Foundation.

Morgan, H.P. 1979. Heat transfer from a buoyant smokelayer beneath a ceiling to a sprinkler spray. 1. A tenta-tive theory. Fire and Materials 3: 27-32.

Morgan, H.P. 1998. Sprinklers and fire safety design. FireSafety Engineering 5: 16-21.

Morgan, H.P., and K. Baines. 1979. Heat transfer from abuoyant smoke layer beneath a ceiling to a sprinklerspray. 2. An experiment. Fire and Materials 3: 34-38.

Morgan, H.P., and J.P. Gardiner. 1990. Design principles forsmoke ventilation in enclosed shopping centres. BR186. Garston, U.K.: Building Research Establishment.

NFPA. 1991. NFPA 910, Recommended practice for the pro-tection of libraries and library collections. Quincy,Mass.: National Fire Protection Association.

DISCUSSIONKenneth Elovitz, Foxboro, Mass.: The paper shows thatsmoke from a moderately large sprinklered fire (up to almost1 MW) will be cool and, therefore, nonbouyant. Why, then,could this paper not be considered an argument in favor ofusing cold smoke to test smoke control systems? Why is thispaper not a contradiction to ASHRAE/NFPA published docu-ments that say cold smoke from smoke bombs does not repre-sent real smoke and is not a proper test medium for smokecontrol systems?Gary D. Lougheed: Atrium smoke management usingmechanical exhaust systems, such as those described in theNFPA/ASHRAE engineering guidelines, are designed assum-ing that the fire is located in the atrium. With the high ceilingheights typical of atria, it is assumed that sprinklers installedin an atrium will not activate until the fire is relatively large.The atrium smoke management system is intended to main-NFPA. 1995. NFPA 92B, Guide for smoke management sys-tems in malls, atria, and large areas. Quincy, Mass.:National Fire Protection Association.

NFPA. 1996. NFPA 13, Standard for the installation ofsprinkler systems. Quincy, Mass.: National Fire Protec-tion Association.

ONeill, J.G., W.D. Hayes, and R.H. Zile. 1980. Full-scalefire tests with automatic sprinklers in a patient room,Phase II. NBSIR 80-2097. Gaithersburg, Md.: NationalInstitute of Standards and Technology.

You, H.-Z., H.-C. Kung, and Z. Han. 1986. Spray cooling inroom fires. NBS-GCR-86-515. Gaithersburg, Md.:National Institute of Standards and Technology.

You, H.-Z., H.-C. Kung, H.-C., and Z. Han. 1989. Theeffects of spray cooling on the ceiling gas temperature atthe door opening of room fires. Fire Safety Science, Pro-ceedings of the Second International Symposium, Hemi-sphere, NY, p. 655-665.ASHRAE Transactions: Symposiatain the smoke height above egress routes to allow buildingoccupants to safely evacuate. The smoke produced by thesefires is buoyant and is not well simulated using the cold smokeproduced using smoke bombs.

The research, upon which the paper Smoke Movementfor Sprinklered Fires is based, is directed at assessing thepotential hazard associated with smoke from a sprinklered firein a mercantile space adjacent to an atrium. Depending on thestage of the sprinkered fire, either buoyant or non-buoyantsmoke will be produced. A mechanical atrium smoke exhaustsystem will provide additional fire protection for the stageduring which the fire produces buoyant smoke. The mainquestions still to be answered using the research are whetheror not the non-buoyant smoke poses a potential hazard and ifthe sprinkler system provides effective smoke managementfor fires in adjacent space during this stage of a fire. Thisresearch does not address methods to test atrium smokeexhaust systems.619

43138.pdfABSTRACTINTRODUCTIONREVIEW OF SPRINKLERED FIRESDESCRIPTION OF EXPERIMENTFigure 1 Large-scale atrium facility and instrumentation (plan).Figure 2 Large scale atrium facility and instrumentation (elevation).Figure 3 Test room and instrumentation (plan).Figure 4 Test room and instrumentation (elevation).PROPANE BURNER TEST RESULTSFigure 5 Temperature profiles for thermocouples in the compartment opening for test with heat rel...Figure 6 Temperature profiles at three locations in the compartment opening for tests at 500 kW.Figure 7 Average temperature profiles in opening for tests at 250 kW.Figure 8 Average temperature profiles in opening for tests at 1000 kW.Figure 9 Average temperature profiles in opening for tests at 2000 kW.Figure 10 Temperature rise vs. heat release rate with 4.1 (L/min)/m2 sprinkler.Figure 11 Temperature rise vs. heat release rate with 8.1 (L/min)/m2 sprinkler.Figure 12 Temperature profiles in opening for tests at 500 kW with opposed air flow.CONCLUSIONSREFERENCESDiscussion